Brain tumors, both benign and malignant, are a major hazard to human health because they impede normal brain function. Traditional detection approaches, such as MRI scans and biopsies, are effective but typically inefficient, especially given the growing patient population. This study uses deep learning approaches to address the demand for an automated, low-cost diagnostic system for brain tumor identification and classification. Two models were created: the first, a convolutional neural network (CNN), which uses binary classification to detect the existence of a tumor. The second model, which combines transfer learning with ResNet-50, classifies detected tumors into four types: glioma, meningioma, pituitary tumor, and no tumor. Both models were highly accurate, with the binary model obtaining 99% accuracy in tumor detection. This dual-model method increases the efficiency and precision of brain tumor diagnosis, making it a feasible clinical solution in places with limited access to professional healthcare.